PROJECT SUMMARY Metastasis, the final biological stage of cancer, is responsible for the majority of cancer-related deaths. With each cancer type spreading to a small set of sites, we know that metastasis is not a random process. However, even tumors of the same type significantly differ in their potential to seed metastases at different sites leading to drastically varying patient survival and potentially sub-optimal treatment. Currently, we cannot accurately predict whether a specific patient?s cancer will become metastatic or not. Only a fraction of patients who receive toxic and expensive therapies benefit from it ? but we do not know how to identify this fraction. We therefore face multiple unmet scientific and clinical challenges in cancer research that can only be overcome by determining the evolutionary rules governing metastatic progression of individual cancers. By utilizing reconstructed cancer phylogenies, we recently showed that some colorectal cancer patients exhibit common origin of metastasis while others exhibit multiple distinct origin of metastasis. Preliminary analysis indicates that phylogenies and the roots of metastasis can be utilized to stratify patients. To test this hypothesis, we propose the following three specific aims: i) perform comprehensive in-silico benchmarking based on established population genetics models across eight methods to robustly infer the roots of spreading metastatic clones, ii) uniformly infer metastatic seeding patterns on cohorts of 49 pancreatic and 17 colorectal cancer patients (528 tumor samples) to determine the predictive power of cancer phylogenies and to quantify the topological distribution of metastases within each patient, and iii) develop mathematical models to characterize the consequences of distinct modes and tempos of dissemination and colonization and thereby provide a quantitative framework to contextualize the observed metastatic seeding patterns. Preliminary calculations show highly non-random patterns suggesting that some subpopulations in the primary tumor have drastically increased metastatic capacity. My long-term goal is to identify and quantify the evolutionary patterns of cancer to improve patient prognosis by predicting metastatic potential and provide desperately needed clinically-actionable information to the physicians for a personalized treatment plan. In addition to the important scientific goals of this Pathway to Independence award, we have developed a curriculum targeting areas in which I would highly benefit from more in-depth training and mentoring before becoming an independent investigator. We therefore propose a series of training activities during the mentored phase to gain experience in a translational biomedical environment and to grow my interdisciplinary skill set, particularly in emerging areas of large-scale biomedical data analysis. These activities coupled with the proposed research will facilitate my transition to independence and will provide a strong foundation to start my own laboratory and write an independent R01 proposal during the R00 phase.